Taming Unleashed Large Language Models With Blockchain for Massive Personalized Reliable Healthcare.

Area of Science:

• Digital Health
• Artificial Intelligence
• Blockchain Technology

Background:

• The digital health field aims for massive, personalized healthcare but faces limitations in clinician resources.
• Large language models (LLMs) have advanced digital health but introduce hallucinations, impacting patient safety.
• Reliable AI integration is crucial for scalable, personalized healthcare solutions.

Purpose of the Study:

• To propose a novel digital hospital system, the Chat Chain-Brain-based Doctor (CHATCBD), for massive, personalized, and reliable healthcare.
• To address LLM-induced hallucinations and enhance patient safety in digital healthcare applications.
• To improve the reliability and trustworthiness of AI-driven diagnostic processes in healthcare.

Main Methods:

• Developed a digital hospital architecture transforming LLM-based diagnostics.
• Designed a controllable AI agents framework with a self-audit mechanism for reliability.
• Integrated blockchain technology for decentralized regulation, personalized routing, and a mathematical model-based audit framework to mitigate LLM hallucinations.

Main Results:

• The CHATCBD system demonstrated significant enhancement of LLM capabilities in personalized healthcare.
• Experiments on 13 datasets confirmed the system's effectiveness from multiple perspectives.
• The blockchain-based audit framework ensured professionalism and honesty in AI-driven medical audits.

Conclusions:

• The proposed CHATCBD system offers a reliable solution for massive, personalized digital healthcare.
• Blockchain integration effectively addresses LLM hallucinations and improves patient-centered decision-making.
• CHATCBD enhances the safety and capabilities of LLMs in personalized healthcare applications.